key: cord-0281443-ml932kp2
authors: Gillich, Astrid; St. Julien, Krystal R.; Brownfield, Douglas G.; Travaglini, Kyle J.; Metzger, Ross J.; Krasnow, Mark A.
title: Alveoli form directly by budding led by a single epithelial cell
date: 2021-12-26
journal: bioRxiv
DOI: 10.1101/2021.12.25.474174
sha: d43726b7737d4faafef310ab0625d2b7751dce93
doc_id: 281443
cord_uid: ml932kp2

Oxygen passes along the ramifying branches of the lung’s bronchial tree and enters the blood through millions of tiny, thin-walled gas exchange sacs called alveoli. Classical histological studies have suggested that alveoli arise late in development by a septation process that subdivides large air sacs into smaller compartments. Although a critical role has been proposed for contractile myofibroblasts, the mechanism of alveolar patterning and morphogenesis is not well understood. Here we present the three-dimensional cellular structure of alveoli, and show using single-cell labeling and deep imaging that an alveolus in the mouse lung is composed of just 2 epithelial cells and a total of a dozen cells of 7 different types, each with a remarkable, distinctive structure. By mapping alveolar development at cellular resolution at a specific position in the branch lineage, we find that alveoli form surprisingly early by direct budding of epithelial cells out from the airway stalk between enwrapping smooth muscle cells that rearrange into a ring of 3-5 myofibroblasts at the alveolar base. These alveolar entrance myofibroblasts are anatomically and developmentally distinct from myofibroblasts that form the thin fiber partitions of alveolar complexes (‘partitioning’ myofibroblasts). The nascent alveolar bud is led by a single alveolar type 2 (AT2) cell following selection from epithelial progenitors; a lateral inhibitory signal transduced by Notch ensures selection of only one cell so its trailing neighbor acquires AT1 fate and flattens into the cup-shaped wall of the alveolus. Our analysis suggests an elegant new model of alveolar patterning and formation that provides the foundation for understanding the cellular and molecular basis of alveolar diseases and regeneration. One Sentence Summary We report a direct budding mechanism of alveolar development distinct from the classical model of subdivision (‘septation’) of large air sacs.

The respiratory surface of the mammalian lung is formed by millions of regularly spaced, densely packed, thin-walled air sacs called alveoli arranged as single units or in small groups along the tubular walls or ends of airways (1, 2). Air passes along the airways into alveoli, where oxygen diffuses across the alveolar wall to reach circulating red blood cells, which distribute it throughout the body (3) . Classic studies of lung structure revealed the geometry and architecture of respiratory airways (1, 2, [4] [5] [6] [7] [8] and provided stereological estimates of alveolar number (2) (3) (4) million in mice), size (50 µm mean diameter), and surface area (6 x 10 3 µm 2 ) (9-12). However, the extreme thinness of alveolar cells and dense packing of alveoli make it difficult to count individual cells and define their structures and boundaries on histological sections. Electron microscopy and serial section reconstruction resolved the fine structure of the air-blood barrier and the organization of the alveolar epithelium (13) (14) (15) (16) (17) (18) , but the techniques are laborious and have not been used to systematically define the morphologies, numbers and arrangement of the different cell types in an alveolus. Hence the three-dimensional cellular structure of alveoli has not been resolved.

There is an urgent need to define the cellular structure and formation of alveoli since they are the sites of major, life-threatening lung diseases. These include chronic diseases such as emphysema/chronic obstructive pulmonary disease (COPD), idiopathic pulmonary fibrosis (IPF), and bronchopulmonary dysplasia (BPD), characterized by impaired or arrested alveolar development, as well as the acute respiratory distress syndromes accompanying severe injury or alveolar damage following infection, as in SARS, MERS, and the current Covid-19 pandemic (19) (20) (21) (22) (23) (24) . In these diseases enlargement, destruction, or flooding of alveoli, or changes in thickness or composition of their walls, result in a loss or lack of gas exchange surface, causing a decline in lung function or even respiratory failure.

The textbook model of alveolar development posits that the epithelium forms large sacs ('sacculation') that are later subdivided into alveoli by contractile myofibroblasts ('septation') Here we use mosaic labeling, clonal analysis and deep imaging to elucidate the threedimensional cellular structure of alveoli in the mouse lung. We then analyze alveolar development at single-cell resolution and at a specific position in the branch lineage. We show that alveoli form much earlier and more simply than previously thoughtby budding of the epithelium on airway stalks between smooth muscle cells that rearrange to become myofibroblasts at the alveolar openings. A single AT2 cell is selected to lead the alveolar outgrowth; a lateral inhibitory signal mediated by Notch prevents induction of AT2 fate in surrounding progenitors, which then develop as AT1 cells. Our analysis suggests an elegant new mechanism of alveolar patterning and development distinct from the classical septation model. S2C ). These classes allowed us to discern two types of alveoli: simple alveoli with the entrance formed by a ring of thick fibers (Fig. 1, A and B, Fig. S2D , and Movie S1), and groups of 2-8 alveoli (alveolar complexes) with a shared entrance encircled by thick fibers and partitions formed by thin fibers (Fig. 1, C and D, Fig. S2E , and Movie S2). This is consistent with previous descriptions of the geometry of respiratory airways (1, 2, 8, 51) . Single alveolar units and complexes are intermingled throughout respiratory airways, although simple outpocketings are especially obvious and may be more abundant on early generations (Fig. S2 , F and G).

To elucidate the cellular structure of alveoli, we determined the morphologies of alveolar cells and their arrangement by labeling single cells using genetic strategies and analyzing their position within alveoli relative to the ring of elastic fibers at the alveolar entrance. Tamoxifeninducible CreER lines (Table S1 ), in which Cre recombinase is fused to a modified estrogen receptor to control the timing and extent of labeling, were crossed to a Cre reporter in which recombination of a loxP flanked stop cassette leads to expression of a fluorescent protein.

Limiting doses of tamoxifen were used to induce rare recombination events (100-800 events per lobe) that labeled isolated cells. Three-dimensional reconstruction of single cells and the surrounding fibers revealed that alveolar cells have a stunning variety of shapes and unusual features and arrangements (Fig. 1, E 

There is on average just a single alveolar type 1 (AT1) cell per alveolus (Fig. 1E, Fig. S3 , and Fig. S5, A, B and G) . It is typically a cup-shaped cell with its thin, expansive cytoplasmic extensions covering nearly the entire alveolar surface (5,500 µm 2 ) (12). However, the arrangement of AT1 cells in alveoli is variable and individual cells can span 2-5 alveoli ( Fig. S3 and S4), consistent with descriptions of 'non-nucleated plates' (17). The majority (96%, n=100) of AT1 cells have one or more (up to 9) holes ('pores of Kohn' (52)) of circular or oval shape and varying sizes (Fig. S5B ). AT1 cells form junctions with each other and with neighboring AT2 cells that are strategically positioned between alveoli as single cuboidal or elongated cells Myofibroblasts, by contrast, are simple, spindle-shaped cells with 2 or 3 cell processes that are aligned with thick elastic fibers at the alveolar entrance ( Fig. 1K and Fig. S5D ). Three to five myofibroblasts form a ring around the alveolar entrance (Fig. S5G) , with their cell bodies located at the intersections of neighboring alveoli and with the number of cells per ring corresponding to the number of intersections.

These clonal labeling experiments combined with immunostaining against common antigens, deep imaging, 3D reconstruction and quantification (Figs. S3-S6 and Methods) allowed us to define the three-dimensional cellular structure of an alveolus ( Fig. 1L and Fig. S5G) . A simple alveolus is composed of just 2 epithelial cells, a single AT2 cell and an AT1 cell that covers the majority of the alveolar surface, 3-4 capillary endothelial cells (of 2 cell types: aerocytes and general capillary cells; relative abundance 1:4), and 5-9 stromal cells including 1-2 pericytes, 1-2 fibroblasts, and 3-5 myofibroblasts. Thus, an alveolus in the mouse lung is composed of just 10-15 cells of 7 major types, excluding immune cells. This is 2-3 fold fewer cells than previous estimates based on stereology of thin sections of alveoli (55) , which apparently did not distinguish complexes from individual alveoli. Alveolar complexes contain proportionally more cells than simple alveoli, scaled to the number of alveolar units ("chambers") in the complex (Fig. S4 ).

To elucidate the cellular mechanisms that initiate and control alveolar patterning, we analyzed the earliest events in alveolar development by mapping the process at a specific position in the branch lineage (56) . In the left primary bronchus (L) lineage, the third anterior branch (A3) forms by domain branching on the anterior aspect of the first secondary branch (L1) and generates 4-6 generations of daughter branches by planar and orthogonal bifurcations ( Fig.   2A ). By immunostaining of whole lobes for Sox2 (or incubation with fluorescent streptavidin to detect biotin expressed by epithelial cells), which allows visualization of conducting airways (57, 58) , we mapped the boundary between conducting and respiratory airways on L.L1.A3 and analyzed when the boundary forms during development (Fig. S7 ). We found that the boundary is positioned on L.L1.A3 daughters at or just distal to the bifurcation between airway generations 4 and 5 (Fig. S7 , A to C). Consistent with previous studies (57) , the boundary is set by embryonic day (E) 16.5 (Fig. 2 , B to D, and Fig. S7D ), and branches distal to the boundary form respiratory airways.

Immunostaining of whole lobes for epithelial and smooth muscle markers showed that at E16.5 smooth muscle actin (SMA)-positive cells surround epithelial tubes destined to form respiratory airways (Fig. 2C ). The cells have an elongated morphology and circumferential orientation typical of smooth muscle on conducting airways (Fig. 2E ). Just one day later (E17.5), the smooth muscle pattern is less organized (Fig. 2D) , with the cells oriented along airway stalks at various angles (Fig. 2F ). Already at this stage, the cells are aligned with elastic fibers (Fig.   S8 ). This suggested that smooth muscle cells on embryonic airways rearrange and become myofibroblasts at the alveolar entrance.

To test whether airway smooth muscle is a source of myofibroblasts, we used a genetic strategy to specifically label and trace smooth muscle on embryonic airways (59) . We combined the airway smooth muscle-specific Lgr6-EGFP-IRES-CreERT2 (Lgr6-CreER) (60) knock-in allele with the Cre reporter Rosa26-tdTomato (61) and induced recombination at E14.5 with a saturating dose of tamoxifen ( Fig. 3A and Fig. S9 ). Analysis of the labeling on postnatal day 10 (P10) showed lineage-labeled cells on first and second order respiratory airways (Fig. 3B) . The cells express SMA and are aligned with thick elastic fibers at alveolar openings ( Fig. 3B') , demonstrating that airway smooth muscle cells become entrance myofibroblasts. Alveolar entrance myofibroblasts express Eln (tropoelastin) as shown by single-cell RNA-sequencing (scRNAseq) and single-molecule in situ hybridization (smFISH) (Fig. S10 , F and G) (62, 63) , confirming that the cells are indeed a source of elastin, a major component of elastic fibers.

To determine if myofibroblasts on later-generation branches also arise from the Lgr6 lineage, we induced recombination at a later embryonic timepoint by dosing E17.5 Lgr6-CreER;

Rosa26-tdTomato lungs with tamoxifen (Fig. S10A ). Lineage labeled cells were located on 3-4 generations of respiratory airways (Fig. S10, B To probe the potential and behavior of individual airway smooth muscle cells, we sparsely labeled smooth muscle cells by induction of rare recombination events in smooth muscle myosin heavy chain (SMMHC)-CreER; Rosa26-Rainbow lungs (64, 65) using limiting doses of tamoxifen (Fig. 3D ). Airway clones induced at E15 and analyzed on postnatal day 10 were composed of 1 to 17 cells. Clones fell into 3 groups based on the location of daughter cells (Table S2 ). In the first group, cells of a clone were located exclusively around stalks of conducting airways ("airway smooth muscle clones", n=30 clones; 44%; Fig. S11 ). In some of these clones (37%) the daughter cells remained in close contact with their siblings, whereas in other clones (37%) cells were dispersed, spanning 100-500 µm along the airway axis, implying extensive movement of daughter cells, as has been observed early in development (59) . The second group was formed by clones spanning the boundary between conducting and respiratory airways ("mixed/boundary clones", n=8 clones; 12%; Fig. 3E and Fig. S12 ). These clones were composed of both elongated cells with circumferential orientation around airway stalks (airway smooth muscle) and spindle-shaped cells aligned with elastic fibers at the alveolar entrance (myofibroblasts). This demonstrates that individual cells can give rise to both, establishing a lineage relationship between airway smooth muscle cells and myofibroblasts. The third group of clones was composed of cells located exclusively on respiratory airways ("myofibroblast clones", n=30 clones; 44%; Fig. 3F and Fig. S13 ). Nearly all (98%) labeled cells (n=100 cells scored from 30 clones) were aligned with alveolar elastic fiber rings. Although the majority (70%) of myofibroblast clones were composed of more than one cell (Table S2) , individual entrance rings were generally incompletely labeled, excluding a model of entrance ring formation by clonal proliferation. The differences in cell behavior our analysis uncovered, with only a fraction of labeled cells turning into myofibroblasts, may reflect differences in the potential of individual cells, or perhaps more likely be dictated by the environment (e.g., only daughter cells located beyond the boundary turn into myofibroblasts).

Immunostaining and smFISH in postnatal (P10) lungs revealed two distinct types of myofibroblasts. Both types express SMA and the myofibroblast marker Fgf18 (66), but one population associates with thick elastic fibers that encircle the openings to simple alveoli and alveolar complexes ("entrance myofibroblasts"; Fig. 4 , A, A', B, B' and C). Our embryonic and postnatal pulse-chase experiments with Lgr6-CreER; Rosa26-tdTomato showed that entrance myofibroblasts derive from the Lgr6 lineage ( Fig. 3B' and Fig. S10 , A to D) and persist in the adult lung ( Fig. 3C and Fig. S10E ). Although nearly all (98%) of these cells turn off SMA, they remain aligned with thick elastic fibers and retain a simple morphology (Fig. 3C') .

The second population also expresses SMA and the myofibroblast marker Fgf18, but these cells are aligned with thin elastic fibers that subdivide alveolar complexes ("partitioning myofibroblasts", Fig. 4 , A, A'', B, B'' and C). Partitioning myofibroblasts also have a spindle shape, and their cell processes are anchored on the thick elastic fibers and entrance myofibroblasts at the openings to alveolar complexes. Unlike alveolar entrance myofibroblasts, partitioning myofibroblasts are not lineage labeled by Lgr6-CreER (Fig. 3B' and Fig. S10 , A to D), hence they arise from another source.

These results suggest that there are two types of myofibroblasts with distinct anatomical locations and developmental origins: SMA+ Fgf18+ Lgr6 lineage-derived entrance myofibroblasts that form the thick fiber rings at the base of simple alveoli and alveolar complexes, and SMA+ Fgf18+ Lgr6 lineage-negative partitioning myofibroblasts that subdivide alveolar complexes. Simple alveoli have entrance but not partitioning myofibroblasts (Fig. 4C ).

The alignment of SMA+ cells with thick elastic fiber rings on stalks of embryonic airways at E17.5 (Fig. S8 ) suggested that alveoli form much earlier than previously assumed (P4 in mice) (25) and by a mechanism distinct from septation.

To determine how alveoli arise on airways, we analyzed the epithelial pattern on stalks of L.L1.A3 daughter branches by staining embryonic (E17.5) lobes for E-cadherin, and also for SMA and elastin (detected by hydrazide) to visualize the arrangement of epithelial cells relative to smooth muscle and elastic fibers. Surprisingly, we found small groups of 1-3 cuboidal epithelial cells protruding between smooth muscle cells on airway stalks ( Fig. 5A and Fig. S14 ).

The budding single cuboidal cells were surrounded by rings of 3 or more smooth muscle cells aligned with elastic fibers (Fig. 5 , A' and A''), indicating that these are the nascent entrance myofibroblasts organizing around budding epithelial cells and depositing elastin, supporting a model of alveolar development by direct budding.

Immunostaining for alveolar epithelial markers, including FGF receptor 2 (Fgfr2), showed that the single cuboidal cells are nascent AT2 cells ( Fig. 5B and Fig. S15 ; see also Brownfield et al., in review (67)). Each budding AT2 cell is surrounded by 2-4 nascent AT1 cells that have begun to flatten (Fig. 5C ). This suggests that the alveolar buds are led by single AT2 cells through the nascent entrance myofibroblasts that begin to synthesize elastic fibers and demarcate the alveolar entrance as the AT2 cell buds ( Fig. 5D and E) .

The regular salt-and-pepper pattern of the alveolar epithelium with single AT2 cells surrounded by 2-4 nascent AT1 cells implies that the alveolar pre-pattern and spacing is established surprisingly early in the process, at or before E17.5. The salt-and-pepper pattern also raised the possibility that the balance between the alternative epithelial fates could be controlled by a Notch-mediated lateral inhibition mechanism (68) . Indeed, an increase in AT2 cells has been observed in embryonic lungs mutant for Lunatic Fringe, a glycosyltransferase that facilitates Notch activation (69) , and Notch signaling is active during the AT2-to-AT1 cell transition in culture (70) . To determine if the Notch pathway is active in developing alveolar epithelial cells, we examined expression of the major transcriptional effector of Notch signaling, CBF1 (RBP-J), in embryonic (E16.5-18.5) lungs using a CBF1-H2B-Venus reporter (71) .

Immunostaining showed nuclear Venus localization in nascent AT1 cells (and endothelial plexus) but little or none in alveolar epithelial progenitors or budding AT2 cells marked by Fgfr2 To test if Notch signaling plays a role in alveolar patterning, we inhibited the pathway in developing (E16-E18.5) lungs using the γ-secretase inhibitor DBZ, which prevents ligandinduced proteolytic processing and activation of Notch (72) . Whereas the majority (88 ± 5%) of AT2 cells in control lungs (n=500 AT2 cells scored in 6 animals) were found as single (isolated) AT2 cells at E18.5, only rare AT2 cells (11 ± 4%) were selected as single cells in DBZ-treated lungs (n= 500 cells in 6 animals), with the rest (89 ± 4%) found in clusters (Fig. 6 , B to D). To determine if Notch inhibition enhances AT2 cell proliferation, as observed at postnatal stages (73) , we performed immunostaining for Ki67. AT2 cell proliferation was comparable in DBZtreated and control lungs (1.2 ± 0.6% Ki67+ AT2 cells in DBZ-treated lungs; 1.5 ± 0.5% in controls; Fig. 6E and Fig. S16E ). This suggests that Notch signaling is required for alveolar patterning and specifically for selection of single AT2 cells, rather than to regulate their proliferation. Technical, developmental timing or background differences could explain the discrepancy between our findings and a previous study of Notch function (73) .

To identify the relevant receptor and ligand, we analyzed expression of Notch pathway components in scRNAseq data for developing mouse lung (47, 74) . This confirmed Notch cells from 58 ± 4% in controls to 87 ± 3% (n= 500 cells scored in 3 experiments). Similarly, blocking antibodies against Notch1 and Notch2 (75) increased AT2 cell selection to 72 ± 7%, and 88 ± 5%, respectively. These results support a role for Notch signaling mediated by Notch1 and Notch2 in alveolar epithelial patterning and lateral inhibition of AT2 fate (Fig. 6H) .

However, the culture experiments do not exclude the possibility that other (non-epithelial) sources could also contribute to alveolar epithelial patterning in vivo.

Our analysis of alveolar patterning and morphogenesis reveals a new mechanism by which alveoli form in the lung. In contrast to the classical ('septation') model in which the epithelium gives rise to large sacs that are later subdivided into alveoli by myofibroblasts, we show that alveoli form early and directly by budding and outgrowth of the epithelium from airway stalks between enwrapping smooth muscle cells (Fig. 7) . These smooth muscle cells rearrange to become a ring of myofibroblasts at the alveolar entrance. This surprisingly simple and elegant way to form alveoli relies on precise spatial and temporal control of epithelial patterning and coordination with surrounding cells.

We propose that the initiating event in alveolar development is the selection of a single AT2 cell that first becomes apparent by Fgfr2 restriction in the embryonic lung (see also Brownfield et al., in review (67) ) right after the boundary is set between conducting and respiratory airways (57) . The nascent AT2 cell leads the alveolar outgrowth and is the source of one or more signals to surrounding cells: a lateral inhibition signal to neighboring epithelial cells mediated by Notch1 and Notch2 that prevents selection of more than one AT2 cell, and possibly a signal to nearby smooth muscle to rearrange to form entrance myofibroblasts and synthesize elastin. Our data support a mechanism of alveolar patterning analogous to Drosophila tracheal branching morphogenesis (76) where a single cell is selected by Fgf signaling to take the lead position in budding, and Notch signaling prevents selection of additional cells.

While Fgf and Notch signaling together control the selection and spacing of leading AT2 cells, and hence set the alveolar pattern, at least one other signal to the epithelium appears necessary to confer AT1 fate on surrounding cells and generate alveoli because expression of a constitutively-active Notch protein was insufficient to induce AT1 differentiation in cultured alveolar epithelial progenitors (not shown). A mechanical signal may play a role in AT1 specification and initial flattening, as suggested by a recent study (46), and pressure-induced forces may ensure further flattening and postnatal outgrowth of alveoli. Whatever the nature of this signal, it can be provided in cultures of purified alveolar epithelial cells induced with Fgf because such cultures readily form alveolospheres with intermingled AT1 and AT2 cells ( Fig.   6F ; see also Brownfield et al., in review (67) ).

In our model of alveolar development, airway smooth muscle is the long sought source of entrance myofibroblasts. They synthesize elastin and are aligned with thick elastic fibers at the base of simple alveoli and alveolar complexes. Although the cells turn off smooth muscle actin in the postnatal (P12-15) lung, some and possibly all of them continue to be associated with alveolar openingsthey may get anchored to the fibers or adhere to the basement membrane of AT1 or capillary cells through specialized junctional complexes (34). Genetic tools to specifically label and manipulate entrance myofibroblasts are needed to directly test the requirement of the cells in alveolus formation and to elucidate their role in the mature lung.

Identification of the signals that recruit or activate them to make or repair fibers may suggest ways to prevent or reverse alveolar elastin destruction in age-associated and heritable forms of emphysema, which could lead to new strategies to replace damaged or missing alveolar entrance rings and inform treatments.

Our results also identified a second type of alveolar myofibroblast with a distinct anatomical location and developmental origin: 'partitioning' myofibroblasts that subdivide alveolar complexes. Interestingly, two molecular types of alveolar myofibroblasts were recently reported in a preprint investigating the diversity of myofibroblasts by scRNAseq at different stages of embryonic and postnatal lung development (77) . One population expresses Lgr6, Hhip and Cdh4 and persists post-alveologenesis, whereas the other expresses Pdgfra and undergoes developmental apoptosis. The Lgr6-expressing subset may correspond to entrance myofibroblasts that we show arise from the Lgr6 lineage, form the thick elastic fiber rings at the openings of simple and complex alveoli, and persist in the adult lung. The second population may be partitioning myofibroblasts that form thin fiber subdivisions of alveolar complexes.

The timing of alveolar complex formation and recruitment of partitioning myofibroblasts appears to coincide with the classical 'alveolarization' stage in the postnatal mouse lung (P4-P30) (25, 27) . This suggests that these myofibroblasts and the associated thin fiber partitions form after the entrance rings of simple and complex alveoli and by a distinct mechanism, presumably part of septation. Although we limited our in depth analysis of alveolar development to stalks of early forming airways to avoid caveats associated with inflation of air-filled lungs and variability in late forming branches, it will be important to find ways around these biological and technical issues to probe at cellular resolution the origin and dynamics of partitioning myofibroblasts and the steps beyond entrance ring formation that expand and subdivide some alveoli into complexes ("septation"). Such approaches could also elucidate other steps in alveolar growth and maturation, including interlinking of alveoli through transseptal AT2 cells with multiple apical sides (Fig. S5C ) and specialized circular junctions between AT1 cells from adjacent airways to form pores of Kohn (Fig. S5B ).

Our model of alveolar development by direct budding is supported by the threedimensional cellular structure of alveoli we elucidated by single-cell labeling. Remarkably, a simple alveolus in the mouse lung is composed of just 10-15 cells of 7 major cell types (excluding immune cells) including a single ('lead') AT2 cell, a single AT1 cell and 3-5 myofibroblasts at the entrance, which is 2-3 fold less than estimates from counting cells and alveoli on sections (55) . Human alveoli are larger (200-300 µm mean diameter compared to 50 µm in mouse) and are thought to contain at least ten times more cells (12, 16, 55, 78) . Single-cell labeling approaches should now be developed to elucidate the precise number and structure of cells in human alveoli and to determine if the mechanism of alveolar development we uncovered in mice, with selection of single lead AT2 cells as the key initiating and patterning event followed by budding and recruitment of entrance myofibroblasts from smooth muscle, is conserved between species. Based on our analysis, the timing of initiation of alveolar development in mouse and humans may be more similar than previously thought. And because humans also appear to have both simple alveoli and complexes (2), the two types of alveolar myofibroblasts, with their distinct origins, anatomy, and functions, may also be conserved. Our analysis of the structure and development of alveoli provides a basis for such evolutionary comparisons and for investigations of how the structure and cells of alveoli are altered with age and in human diseases that compromise gas exchange, including chronic lung diseases that arise in infants born too early with incompletely developed alveoli (79) . By identifying the cells and molecules that control alveolar patterning and morphogenesis, our analysis suggests ways to stimulate failed or arrested alveolar development in premature lungs, and may inform strategies to rebuild alveoli in injured, aged or diseased lungs.

The following mouse strains were used: C57BL/6 (C57BL/6NCrl, Charles River Laboratories, strain 027) was the wild-type strain. Hopx-CreER (Hopx 

Embryonic mouse lungs were dissected in PBS and fixed in 4% paraformaldehyde (PFA) for 1 hour (E16.5), 1.5 hours (E17.5) or 2 hours (E18.5) at 4°C. Fixed lungs were washed three times in PBS for 15 minutes at 4°C, then dehydrated through a PBS and methanol series into 100% methanol and stored at -20°C. For Sox2 antibody stains, embryonic lungs were fixed in 0.5% PFA for 3 hours at room temperature, followed by washes and dehydration into methanol. For Fgfr2 stains, lungs were incubated in Dent's fixative (80% methanol, 20% dimethylsulfoxide (DMSO)) overnight at 4°C, washed twice in 100% methanol and stored as above.

Postnatal mouse lungs were perfused with PBS, inflated with 2% low-melting point agarose (Thermo Fisher, Cat. No. 16520100) dissolved in PBS and fixed by immersion in 4% PFA for 2 hours (P10) or 3-5 hours (adult) at 4°C (or in 0.5% PFA for 5 hours at room temperature for Sox2 antibody stains). Fixed tissue was washed three times in PBS for 30 minutes and dehydrated in methanol as above. Lungs were rehydrated on the day of staining through a methanol and PBS series into PBS. Sections (300-500 µm) were cut from left or right cranial lobes on a vibratome (Leica Biosystems). Rehydrated whole lobes or vibratome sections were blocked in 5% heat inactivated goat serum (or donkey serum for primaries raised in goat) in PBS with 0.5% Triton X-100 and 3% BSA for 2 hours at room temperature and incubated in primary antibodies diluted in block for 3 nights at 4°C. 

Postnatal lungs, inflated as described above, were fixed in 10% neutral buffered formalin (Fisher Scientific) for 24 hours at room temperature and transferred to 70% ethanol (made up in PBS) following 3 brief washes in PBS for embedding in paraffin. Sections were cut at 6 μm. smFISH was performed using a proprietary high-sensitivity RNA amplification and detection technology (56), which allows analysis of alveolar development without inflation. We analyzed stalks (rather than branch tips) to map the position of epithelial cells and smooth muscle cells relative to the axis of the airway tube.

Lgr6-EGFP-IRES-CreERT2 (Lgr6-CreER) (60) males were bred to Rosa26-tdTomato (61) females to generate heterozygous Lgr6-CreER males homozygous for Rosa26-tdTomato. The males were bred to CD1 wild-type females (Charles River Laboratories), since they were found to tolerate higher doses of tamoxifen than C57BL/6. Pregnant females from these crosses were dosed at E14.5 with saturating (4 mg) tamoxifen to label smooth muscle on embryonic airways with heritable tdTomato expression and the labeling was analyzed at E16.5, E17.5, P10, and P60 

The γ-secretase inhibitor DBZ (Tocris, Cat. No. 4489) was delivered by daily intraperitoneal injection in timed-pregnant (E16-E18) C57BL/6 females at 30 µmol per kilogram body weight. 

Epithelial progenitors were isolated as previously described (40) 

Images were acquired using inverted Zeiss 780 or upright Zeiss 880 confocal laser scanning microscopes. Zen Imaging Software (Zeiss) and Adobe Photoshop were used to adjust image levels and pseudocolor the images. Volocity Software (Perkin Elmer) was used to generate maximum intensity projections from z-stacks. The fine filter was applied to images of Sox2 antibody stained lungs. Imaris Software (Bitplane) was used for 3D visualization.

Processed scRNAseq MARS-Seq data for developing (E16.5, E18.5) mouse lung by Cohen et al. 

Figures S1-S17 Table S1 . Cre and CreER drivers used for clonal labeling and lineage tracing. Movie S1. Three-dimensional architecture of a simple alveolus.

Movie S2. Three-dimensional architecture of an alveolar complex. Arrowheads, compartment boundary (identified by staining with fluorescent streptavidin to detect epithelial biotin; not shown). Brackets, respiratory airways with lineage-labeled cells.

Lineage-labeled cells are located on 2 generations of respiratory airways, which corresponds to the number of airway generations surrounded by smooth muscle beyond the compartment boundary at E16.5 (see Fig. 2C ). Incomplete labeling of alveolar entrance rings on these airways may be due to inefficient recombination in airway smooth muscle (see Fig. S9 ) and partitioning myofibroblasts (pink) that subdivide alveolar complex into two units. Outpocketing to form alveoli is accompanied by flattening and rounding of AT1 cells into a cup shape (one cell highlighted in green) and may be driven by mechanical forces around birth. and diagrams (bottom) as in Figure S3 showing the arrangement of single AT1 cells (green) as See Fig. 6E for quantification of proliferation. Scale bars, 50 µm. Gillich et al., Figure S15 * * * * * *

Pulmonary acinus: geometry and morphometry of the peripheral airway system in rat and rabbit

Morphometry of the human pulmonary acinus

The pathway for oxygen. Structure and function in the mammalian respiratory system

Three-dimensional reconstruction of the rat acinus

The structure of the pulmonary acinus in a child of six years and eight months

The morphology of the finer branches of the bronchial tree of the human lung

Morphology of the acinus of the human lung

Branching pattern of airways and air spaces of a single human terminal bronchiole

The number of alveoli in the human lung

Stereological estimates of alveolar number and size and capillary length and surface area in mice lungs

The three-dimensional ultrastructure of the human alveolar epithelium revealed by focused ion beam electron microscopy

The normal human lung: ultrastructure and morphometric estimation of diffusion capacity

The pulmonary alveolar epithelium of laboratory mammals and man

Morphometric characteristics of cells in the alveolar region of mammalian lungs

Epithelial Vegfa specifies a distinct endothelial population in the mouse lung

Allometric relationships of cell numbers and size in the mammalian lung

The branching programme of mouse lung development

Two nested developmental waves demarcate a compartment boundary in the mouse lung

Tissue optical clearing, threedimensional imaging, and computer morphometry in whole mouse lungs and human airways

Mesenchymal cells. Defining a mesenchymal progenitor niche at single-cell resolution

Lgr6 marks stem cells in the hair follicle that generate all cell lineages of the skin

A robust and high-throughput Cre reporting and characterization system for the whole mouse brain

A molecular cell atlas of the human lung from single-cell RNA sequencing

Alveolar cell fate selection and lifelong maintenance by FGF signaling

The Notch signalling system: recent insights into the complexity of a conserved pathway

Lunatic Fringe-mediated Notch signaling is required for lung alveogenesis

Dlk1-mediated temporal regulation of Notch signaling is required for differentiation of alveolar type II to type I cells during repair

A bright single-cell resolution live imaging reporter of Notch signaling in the mouse

Modulation of Notch processing by gamma-secretase inhibitors causes intestinal goblet cell metaplasia and induction of genes known to specify gut secretory lineage differentiation

Epithelial Notch signaling regulates lung alveolar morphogenesis and airway epithelial integrity

Lung single-cell signaling interaction map reveals basophil role in macrophage imprinting

Therapeutic antibodies reveal Notch control of transdifferentiation in the adult lung

Social interactions among epithelial cells during tracheal branching morphogenesis

Three-axis classification of mouse lung mesenchymal cells reveals two populations of myofibroblasts doi

Cell number and cell characteristics of the normal human lung

Bronchopulmonary dysplasia

We thank C. Siebel 

Gillich et al., Figure S17 ln(UP10K+1)